Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

The crisp autumn air bit at Liam’s cheeks as he stood on the pitcher's mound. The roar of the crowd was a dull hum in the background. He’d worked tirelessly, honing his fastball, his curveball, his slider – perfecting the art of getting batters out. Now the coach was yelling at him from the dugout, "Liam, we need you in right field!" Confusion flickered in his eyes, but quickly transformed into resolve. He strode towards right field, his glove slapping against his thigh. He felt a thrill of anticipation. The open space, the chance to chase down fly balls, the responsibility of covering the outfield – a different kind of thrill than what he was used to.

Emotion: satisfied

Cluster: Positive / Joy
PC1 (Valence): 3.20 Positive
PC2 (Disposition): -1.31

Role in Research

This story is one of 1,000 stories generated for the emotion satisfied. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 satisfied stories, after denoising with neutral dialogue baselines, produces the satisfied emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the satisfied vector is projected through the unembedding matrix.

Promoted:
la0.518
happy0.465
enjoy0.439
relaxed0.430
enjoys0.429
Suppressed:
S-0.515
C-0.511
😣-0.510
worse-0.477
😖-0.453